Tensor to classification🔗

Given a tensor representing classification results, outputs the indices of accepted classes and the corresponding tensor values.

Inputs🔗

  • tensor: An input tensor of confidence values in which at most one dimension differs from one. The tool will reshape the tensor to a 1-by-N matrix.

  • confidenceThreshold: The minimum confidence value that will be accepted. Set to zero to accept all values.

  • maxResultCount: The maximum number of accepted results. Set to one to pick only the entry with the greatest confidence.

  • scaling: The way confidence values in the input tensor are scaled.

Outputs🔗

  • index: The indices of selected entries as a N-by-1 matrix.

  • confidence: The corresponding confidence values in the input tensor, sorted in descending order.

enum Scaling🔗

Scaling performed to the input tensor confidence values.

Values:

enumerator None🔗

The input confidence values are used as such.

enumerator Softmax🔗

The input confidence values are scaled to range [0, 1] using the softmax function.